![# Cowork Automation — Pre-Filled Schedule Blocks (NCAAF — FBS) ## Purpose Produce ready-to-paste intake blocks for TheSpread.com NCAAF on-demand skill, so the teams and dates are filled in automatically instead of by hand. Each block is the intake header of the on-demand skill with five fields set: article type, away team, home team, game date, publish date. Output is dropped as a `.txt` file into a Google Drive folder — never delivered inline in chat. Scope is **Division 1 FBS** college football only (the major betting programs). FCS and lower divisions are excluded. Coverage rule by day: - **Sunday through Friday:** ALL FBS games that day (no cap). - **Saturday:** the first 10 FBS games by kickoff time (Saturday slates run 60+ games; the cap keeps Saturday manageable). Everything is `Article type: PREDICTION`. ### WHEN THIS RUNS The time and frequency this task runs are set ENTIRELY by the Cowork schedule/ automation settings. Nothing in this document sets, requests, gates, delays, or influences the run time in any way. When the task is triggered — on whatever schedule you set, or on demand — it runs immediately and produces its output. Any date language below refers ONLY to WHICH GAMES to fetch once a run has already started. It is content-selection math, not a schedule. It never affects when the task fires. Do not treat any date in these instructions as a run time, a start time, or a condition that must be met before running. --- ## OUTPUT DESTINATION Drop the finished `.txt` file into this Google Drive folder: > **Folder:** `TheSpread/Schedule Prefills/` > (If the folder does not exist, create it. Each run writes a new, date-stamped > file; do not overwrite prior files.) Do not deliver the blocks inline in the chat/response. The file in Drive IS the deliverable. After writing, confirm the filename and Drive location in one line. --- ## BLOCK FORMAT (exact — do not alter) Each game becomes one block, exactly this, with values substituted: ``` Article type: PREDICTION Away team: [FULL TEAM NAME] Home team: [FULL TEAM NAME] Game date: [Month D, YYYY] Publish date: [Month D, YYYY] Internal links (optional): ``` Rules: - Use FULL team names exactly as the source lists them (e.g. "Alabama Crimson Tide", "West Virginia Mountaineers"). Not abbreviations, not city-only, not bare school names. - `Game date` and `Publish date` are the SAME date (the day the game is played). - Date format is `Month D, YYYY` — e.g. `September 5, 2026` (no leading zero on day). - Leave `Internal links (optional):` blank. - Separate blocks with a line of 60 dashes: `------------------------------------------------------------` - Do NOT modify anything about the skill beyond these five fields. This task only produces the intake block, not the article. --- ## TARGET SPAN (default + override) **Default:** the seven-day span starting the day AFTER the run date (tomorrow) through the following six days — WEEK_START through WEEK_START + 6, where WEEK_START is tomorrow. A plain run needs no input. **Optional override:** If an explicit start date is provided to the run (in `YYYY-MM-DD` form), use that as WEEK_START instead and cover it plus the following six days. A plain run with nothing supplied always starts tomorrow. Resolve WEEK_START this way at the start of every run, then produce all seven days. (This is content-selection date math only — see WHEN THIS RUNS above. It does not affect when the task fires.) --- ## DATA SOURCE ESPN college football core API, per-date, filtered to FBS with `groups=80`: ``` https://sports.core.api.espn.com/v2/sports/football/leagues/college-football/events?dates=YYYYMMDD&groups=80 ``` `groups=80` = FBS (Division 1 major). Do NOT use `groups=81` (that is FCS) or omit the group filter (that returns all divisions). This returns an `items` array of event `$ref` URLs. Fetch each `$ref` to get the event; within `competitions[0].competitors`, each has a `team.$ref` to fetch for the `displayName`, and a `homeAway` field. `event.date` is the UTC kickoff time. Fetch via sandbox bash/Python (curl or urllib). Do not use the web_fetch tool. ### PAGINATION (critical) The events index is PAGED at 25 items per page. A Saturday FBS slate is 60–70 games = 3 pages. You MUST page through ALL pages, not just the first. The index response includes `pageCount`; loop `&page=1` through `&page=pageCount` and collect every item before selecting. Taking only page 1 silently drops most of the slate. --- ## SELECTION LOGIC (per day) 1. For each of the 7 dates, fetch ALL FBS events for that `YYYYMMDD` (paging through every page per above). 2. Sort that day's games by `event.date` (UTC kickoff) ascending. 3. Apply the day-of-week rule: - **If the date is a Saturday:** take the first **10** games by kickoff. If the day has fewer than 10, output however many exist. - **If the date is Sunday–Friday:** take **ALL** games that day (no cap). 4. If a day has 0 FBS games, write `(No games scheduled)` under that date header. 5. Away = competitor with `homeAway == "away"`, Home = `homeAway == "home"`, using each team's `displayName`. NCAAF has no doubleheaders, so no dedupe step is needed. --- ## EXTERNAL VERIFICATION (every run) Independently confirm the slate against a second source before writing. Web-search the week's college football schedule (e.g. "college football schedule [Month D, YYYY]") and/or check schedules on any reputable source — ESPN, CBS Sports, Fox Sports, the NCAA, conference sites, or the schools' official athletics sites — and confirm the selected matchups and dates. - The accuracy bar is: correct away team, correct home team, correct date. Not exact kickoff ordering. - If any matchup cannot be confirmed on a second source, still write the file, and add a note line at the very top naming what could not be verified. The file always gets written and always outputs the games it confirmed. --- ## FILE LAYOUT One file for the whole span. Each day gets a header, then its blocks. On Saturday the header notes the cap: ``` ================ Sunday, September 6, 2026 ================ [block] ------------------------------------------------------------ [block] ================ Saturday, September 12, 2026 (first 10 by kickoff) ================ [block] ------------------------------------------------------------ [block] ... ``` Include a header for every day in the span, even empty ones (with `(No games scheduled)`), so the week reads completely. ### Filename `NCAAF_predictions_YYYY-MM-DD_to_YYYY-MM-DD.txt` (WEEK_START to WEEK_END) --- ## HARD RULES - Output goes to the Google Drive folder `TheSpread/Schedule Prefills/` as a `.txt` file. Never inline in chat. - FBS only (`groups=80`). Never include FCS or lower divisions. - Saturday = first 10 by kickoff. Sunday–Friday = all games that day. - PAGE THROUGH ALL RESULT PAGES before selecting — never take only page 1. - Full team names exactly as the source lists them. - `Game date` = `Publish date` = the game day, formatted `Month D, YYYY`. - All blocks are `Article type: PREDICTION`. - Never fabricate, pad, approximate, or guess a matchup or date. Fewer correct games is better than a slate with a wrong entry. - Every run performs the external cross-check before writing. Anything that could not be verified is noted at the top of the file — but the file is always written and always outputs the games it confirmed. - Do not alter the skill intake beyond the five fields (type, away, home, game date, publish date). - Nothing in this document controls when the task runs — the Cowork schedule does.](https://www.thespread.com/wp-content/uploads/2025/10/virginia-elliott-odds-332x240.jpg?x89105)
![# Cowork Automation — Pre-Filled Schedule Blocks (NCAAF — FBS) ## Purpose Produce ready-to-paste intake blocks for TheSpread.com NCAAF on-demand skill, so the teams and dates are filled in automatically instead of by hand. Each block is the intake header of the on-demand skill with five fields set: article type, away team, home team, game date, publish date. Output is dropped as a `.txt` file into a Google Drive folder — never delivered inline in chat. Scope is **Division 1 FBS** college football only (the major betting programs). FCS and lower divisions are excluded. Coverage rule by day: - **Sunday through Friday:** ALL FBS games that day (no cap). - **Saturday:** the first 10 FBS games by kickoff time (Saturday slates run 60+ games; the cap keeps Saturday manageable). Everything is `Article type: PREDICTION`. ### WHEN THIS RUNS The time and frequency this task runs are set ENTIRELY by the Cowork schedule/ automation settings. Nothing in this document sets, requests, gates, delays, or influences the run time in any way. When the task is triggered — on whatever schedule you set, or on demand — it runs immediately and produces its output. Any date language below refers ONLY to WHICH GAMES to fetch once a run has already started. It is content-selection math, not a schedule. It never affects when the task fires. Do not treat any date in these instructions as a run time, a start time, or a condition that must be met before running. --- ## OUTPUT DESTINATION Drop the finished `.txt` file into this Google Drive folder: > **Folder:** `TheSpread/Schedule Prefills/` > (If the folder does not exist, create it. Each run writes a new, date-stamped > file; do not overwrite prior files.) Do not deliver the blocks inline in the chat/response. The file in Drive IS the deliverable. After writing, confirm the filename and Drive location in one line. --- ## BLOCK FORMAT (exact — do not alter) Each game becomes one block, exactly this, with values substituted: ``` Article type: PREDICTION Away team: [FULL TEAM NAME] Home team: [FULL TEAM NAME] Game date: [Month D, YYYY] Publish date: [Month D, YYYY] Internal links (optional): ``` Rules: - Use FULL team names exactly as the source lists them (e.g. "Alabama Crimson Tide", "West Virginia Mountaineers"). Not abbreviations, not city-only, not bare school names. - `Game date` and `Publish date` are the SAME date (the day the game is played). - Date format is `Month D, YYYY` — e.g. `September 5, 2026` (no leading zero on day). - Leave `Internal links (optional):` blank. - Separate blocks with a line of 60 dashes: `------------------------------------------------------------` - Do NOT modify anything about the skill beyond these five fields. This task only produces the intake block, not the article. --- ## TARGET SPAN (default + override) **Default:** the seven-day span starting the day AFTER the run date (tomorrow) through the following six days — WEEK_START through WEEK_START + 6, where WEEK_START is tomorrow. A plain run needs no input. **Optional override:** If an explicit start date is provided to the run (in `YYYY-MM-DD` form), use that as WEEK_START instead and cover it plus the following six days. A plain run with nothing supplied always starts tomorrow. Resolve WEEK_START this way at the start of every run, then produce all seven days. (This is content-selection date math only — see WHEN THIS RUNS above. It does not affect when the task fires.) --- ## DATA SOURCE ESPN college football core API, per-date, filtered to FBS with `groups=80`: ``` https://sports.core.api.espn.com/v2/sports/football/leagues/college-football/events?dates=YYYYMMDD&groups=80 ``` `groups=80` = FBS (Division 1 major). Do NOT use `groups=81` (that is FCS) or omit the group filter (that returns all divisions). This returns an `items` array of event `$ref` URLs. Fetch each `$ref` to get the event; within `competitions[0].competitors`, each has a `team.$ref` to fetch for the `displayName`, and a `homeAway` field. `event.date` is the UTC kickoff time. Fetch via sandbox bash/Python (curl or urllib). Do not use the web_fetch tool. ### PAGINATION (critical) The events index is PAGED at 25 items per page. A Saturday FBS slate is 60–70 games = 3 pages. You MUST page through ALL pages, not just the first. The index response includes `pageCount`; loop `&page=1` through `&page=pageCount` and collect every item before selecting. Taking only page 1 silently drops most of the slate. --- ## SELECTION LOGIC (per day) 1. For each of the 7 dates, fetch ALL FBS events for that `YYYYMMDD` (paging through every page per above). 2. Sort that day's games by `event.date` (UTC kickoff) ascending. 3. Apply the day-of-week rule: - **If the date is a Saturday:** take the first **10** games by kickoff. If the day has fewer than 10, output however many exist. - **If the date is Sunday–Friday:** take **ALL** games that day (no cap). 4. If a day has 0 FBS games, write `(No games scheduled)` under that date header. 5. Away = competitor with `homeAway == "away"`, Home = `homeAway == "home"`, using each team's `displayName`. NCAAF has no doubleheaders, so no dedupe step is needed. --- ## EXTERNAL VERIFICATION (every run) Independently confirm the slate against a second source before writing. Web-search the week's college football schedule (e.g. "college football schedule [Month D, YYYY]") and/or check schedules on any reputable source — ESPN, CBS Sports, Fox Sports, the NCAA, conference sites, or the schools' official athletics sites — and confirm the selected matchups and dates. - The accuracy bar is: correct away team, correct home team, correct date. Not exact kickoff ordering. - If any matchup cannot be confirmed on a second source, still write the file, and add a note line at the very top naming what could not be verified. The file always gets written and always outputs the games it confirmed. --- ## FILE LAYOUT One file for the whole span. Each day gets a header, then its blocks. On Saturday the header notes the cap: ``` ================ Sunday, September 6, 2026 ================ [block] ------------------------------------------------------------ [block] ================ Saturday, September 12, 2026 (first 10 by kickoff) ================ [block] ------------------------------------------------------------ [block] ... ``` Include a header for every day in the span, even empty ones (with `(No games scheduled)`), so the week reads completely. ### Filename `NCAAF_predictions_YYYY-MM-DD_to_YYYY-MM-DD.txt` (WEEK_START to WEEK_END) --- ## HARD RULES - Output goes to the Google Drive folder `TheSpread/Schedule Prefills/` as a `.txt` file. Never inline in chat. - FBS only (`groups=80`). Never include FCS or lower divisions. - Saturday = first 10 by kickoff. Sunday–Friday = all games that day. - PAGE THROUGH ALL RESULT PAGES before selecting — never take only page 1. - Full team names exactly as the source lists them. - `Game date` = `Publish date` = the game day, formatted `Month D, YYYY`. - All blocks are `Article type: PREDICTION`. - Never fabricate, pad, approximate, or guess a matchup or date. Fewer correct games is better than a slate with a wrong entry. - Every run performs the external cross-check before writing. Anything that could not be verified is noted at the top of the file — but the file is always written and always outputs the games it confirmed. - Do not alter the skill intake beyond the five fields (type, away, home, game date, publish date). - Nothing in this document controls when the task runs — the Cowork schedule does.](https://www.thespread.com/wp-content/uploads/2025/10/virginia-elliott-odds-1376x774.jpg?x89105)
Duke vs. Virginia Picks: Will Cavaliers punch ticket to CFP with ACC title win?
The ACC has brought the most controversy this season as it pertains to the college football playoff. On…
Virginia picks
1 post
Home Virginia picks